Directiva r1.2 design and build

[Alan J]

OK - here are some off-axis measurements.

 

[Alan J]

Here's a quick comparison of vertical directivity at selected angles (modest vertical angles I've been watching for issues when standing, plus high and low bounce angles) for the model vs. breadboarded crossover. I guess the right words are "pretty close." I'd love to do a bit better at +15 degrees (and the real thing has 0.3 dB more of a drop than the model) but I also don't think that I can hear the change.

Model (normalized to on-axis):

Measured (normalized to on-axis)

 

[Alan J]

One addendum/clarification regarding comparisons between VCAD predictions and as-built measurements:

For the comparisons of posts 156 and 162, I used actual component values, as measured using a DATS V3, in VCAD. Actual values included the DATS measurements of inductor series resistance, which was close to expected values, but not capacitor ESR. The ESR values I get from DATS (after calibration) seem way high.

I'd be interested in whether others have gotten ESR values that they trust from a DATS box.

 

[Okankav]

hi alan,

Can you share what type of acoustic materials you used in the cabinet? acoustilux, bitumen pads or felt

 

[Okankav]

also if is possible can u share photos inside the updated box?

 

[Alan J]

All I have inside is an inch of shredded denim batting around the perimeter (top/sides/bottom), with no other fill or damping. I'm traveling for the coming week but hope to jump back in after I return with crossover construction and mounting - and getting the other box caught up - and will definitely post some interior pics.

 

[Alan J]

Since I've been silent for a couple of weeks...

Crossovers are coming along. As expected, I went with separate boards for mid-woofer and tweeter so that inductors can lay flat.

The board is this material which cut and drilled cleanly (a small but not consequential amount of tearout at the bottom of holes if drilled unsupported). It seems like far better quality than hardboard from the big box building supply stores.

Un-soldered spots are where the driver wiring connects. I'll place the Purifi section on the "floor" and the DXT section either on the "ceiling" or on one side depending upon how the wiring lays out.

 

[terryforsythe]

Looks good. If you can mount one crossover on the back panel and one on a side panel, that probably would be optimal - it would place the inductors perpendicular to one another. If not, mount them on opposite side panels.

 

[Alan J]

The first speaker is closed up and ready for revisiting bass alignment.

 

[terryforsythe]

Very nice!

 

[AlmaAtaKZ]

What are the black connecting screw blocks called?

 

[somebodyelse]

What are the black connecting screw blocks called?

They look like a custom 3d printed version of a modesty block

 

[Alan J]

They look like a custom 3d printed version of a modesty block

That's basically what they are, except that the hole for the baffle is elongated to allow for seasonal movement.

I ended up spending a bit of time on this because of the requirements. The baffle mounting screws need to be tight enough to compress the O-ring that provides the seal. I needed to be sure that the blocks wouldn't just bend (or break). The 3D printing layers are perpendicular to the holes for the flat head screws. The two flat head screws into the side walls compress the layers. I did a bunch of tests with different approaches and this one worked well. I snapped some heads off of screws but couldn't break these blocks with this layer orientation. The material is PETG which doesn't creep much.

 

[Alan J]

Turning my attention back to bass alignment...

I obtained my net volume estimate by starting with internal box dimensions, adding volume for cutouts, and subtracting estimates for magnet assemblies, driver & PR cones, etc. For the crossovers I used actual component and board dimensions, then added 15% for wires, tie-downs, etc. My target was 10 liters. The estimate came out to 9.99.

Jeff Bagby's box model spreadsheet has auto-alignment for passive radiators which provided a box volume of 8.35 L and a tuned fb of 40 Hz. With my slightly larger 10 L box, suggested fb is 37 Hz which requires 39 g extra mass on the PR. Of course, there are assumptions being made here regarding Qa and Ql. Series R is the resistance of the crossover inductor.

Here are VCAD predictions of bass responses with different added masses:

I'm partial to responses reminiscent of critically damped sealed boxes, even as I don't think I could give a clear theoretical reason. I definitely don't want "one-note bass". So [EDIT] I've started at 41 g (available washers) and it sounds great on inital listening. The plan is grab near-field data to see how close we are, and of course listen some more.

 

[Alan J]

More reason to be impressed by the tools that are available to us... Here's the VCAD prediction for bass response at 40g added PR mass, compared with merged nearfield-farfield data:

 

[Alan J]

Somewhat embarrassing update... My "final" measurements showed the speaker to be a few dB hot above 5-10 KHz, then a notable rolloff above 10 KHz. With the pair playing in a different room, I think I can hear the hot treble with some source material. I don't think my aging ears would ever hear the rolloff.

Anyway, the problem was a microphone cal file error. The last set of driver measurements were taken with my EMM-6 using the cal for my Umik-1. Lessons: Even though REW keeps track of cal files for different sources (so I don't know what happened), (1) rename cal files to include the mic name, and (2) CHECK EVERYTHING before starting measurements!

I should have noticed this. The error made the DXT seem not to have its normal high-frequency rolloff. One of the folks who DM'd me here expressed surprise at that. I wrongly attributed this to some changes in the facets. With more experience, I would have recognized that those changes would not explain this change in the response.

The good news:

• Measurements can be corrected, as REW allows a cal file change (changing the cal in REW or using the arithmetic tool in VCAD yield the same results).
• In simulation, the high frequency low pass (cap across the tweeter) becomes a modest-Q 10 KHz notch filter. On-axis and listening window responses look great, and the sound power is smoother than before.
• Crossover mods are straightforward, involving only the tweeter, and there's room on the existing board. Parts are to arrive tomorrow.

The plan was to deliver to the intended owners at the end of the week, and that's still my plan.

 

[Alan J]

Here's the original crossover with correct microphone calibration applied. As noted before, I doubt that I personally can hear much of that high frequency dropoff. But I did feel like I could hear some harshness in some material from that low-Q hump around 6 Khz that's most obvious in the sound power response.

A modified crossover for the DXT:

And finally, with trimming of the Purifi mid-band level:

Trimming the Purifi level a bit improves linearity in the listening window, which is set to 0-20 degrees, at the expense of a slight departure of sound power from the -1.2 dB/oct line. I'm going with this option.

 

[Alan J]

The mods are made to the first unit, and I collected a full set of measurements. As shown below, we're now about 2.5 dB above the desired -0.1 dB/octave line at 10 KHz, and there's a sizeable rise in sound power in that region.

I had reason to be fully confident in this crossover change: In the graph below, the orange dashed trace is an actual microphone measurement with the original crossover. The other traces are VCAD results for that same crossover (with properly cal'd driver data). In other words, with the properly calibrated driver data, VCAD's prediction was uncannily close to what I measured. The same ought to be true for the modified crossover.

I've looked everywhere: Re-checked sound card and mic calibrations, checked that the mods are intact (paralleling R4 with a smaller resistor results in a notch around 6-8 KHz as expected), re-checked the corrected calibrations of the driver data. All component values were checked before installation. The mods are clean (photo below). There's good clearance between the new inductor and the PR frame (and higher-frequency response is unaffected by whether the PR is installed).

So I'm kind of at a loss. C3 is a ClarityCap PX series. I thought about the possibility of the cap performing poorly (e.g., a big rise in ESR) at higher frequencies, but that doesn't explain the strong agreement between VCAD and my measurements when that same cap was in the original crossover.

 

[Paweł L]

What the ... That's weird. Any decent film cap has negligible ESR effect at audio frequencies. And I don't think the the amp would roll off the 10k-20k range by that much. But there has to explanation to this behavior. EPDR impedance equivalent drops to 1ohm with your tweeter section, do maybe the amp doesn't like load ‍
Anyway this is the best I could come up with that doesn't drop that low as far as impedance is concerned, and has similar electrical transfer function to your tweeter section.

 

[Alan J]

To be clear (or hoping to be clearer!), what I'm seeing is a low-Q rise, about 2.5 dB at peak, centered around 10 KHz. The rolloff in the first VCAD shot in Post 177 is with the original crossover that had the cap directly across the tweeter (@Paweł L - you'd noted that this was unusual). The graph at the top of Post 178 shows that rolloff replicated in VCAD once the calibration of the driver data was corrected.

How well VCAD predicts that rolloff gave me confidence in both the simulation and the measurements. The simulation for the new crossover, with the same driver data, is in the last panel of Post 177.